Archive of Applied Mechanics 69 (1999) 703±725 Ó Springer-Verlag 1999

Size effect on structural strength: a review

Z. P. BazÏant

703
Summary The article attempts a broad review of the problem of size effect or scaling of failure,
which has recently come to the forefront of attention because of its importance for concrete and
geotechnical engineering, geomechanics, arctic ice engineering, as well as for designing large
load-bearing parts made of advanced ceramics and composites, e.g. for aircraft or ships. First,
the main results of Weibull statistical theory of random strength are brie¯y summarized, and
its applicability and limitations described. In this theory as well as plasticity, elasticity with a
strength limit, and linear elastic fracture mechanics (LEFM), the size effect is a simple power
law, because no characteristic size or length is present. Attention is then focused on the
deterministic size effect in quasibrittle materials which, because of the existence of a non-
negligible material length characterizing the size of the fracture process zone, represents the
bridging between the simple power-law size effects of plasticity and of LEFM. The energetic
theory of quasibrittle size effect in the bridging region is explained, and then a host of recent
re®nements, extensions and rami®cations are discussed. Comments on other types of size
effect, including that which might be associated with the fractal geometry of fracture, are also
made. The historical development of the size-effect theories is outlined, and the recent trends of
research are emphasized.

Key words Scaling, size effect, fracture mechanics, quasibrittle materials, asymptotic methods

1
Introduction
The size effect is a problem of scaling, which is central to every physical theory. In ¯uid
mechanics research, the problem of scaling continuously played a prominent role for over a
hundred years. In solid mechanics research, though, the attention to scaling had many inter-
ruptions and became intense only during the last decade.
Not surprisingly, the modern studies of nonclassical size effect, begun in the 1970's, were
stimulated by the problems of concrete structures, for which there inevitably is a large gap
between the scales of large structures (e.g. dams, reactor containments, bridges) and of labo-
ratory tests. This gap involves in such structures about one order of magnitude; even in the rare
cases when a full-scale test is carried out, it is impossible to acquire a suf®cient statistical basis
on the full scale.
The question of size effect recently became a crucial consideration in the efforts to use
advanced ®ber composites and sandwiches for large ship hulls, bulkheads, decks, stacks and
masts, as well as for large load-bearing fuselage panels. The scaling problems are even greater
in geotechnical engineering, arctic engineering, and geomechanics. In analyzing the safety of an
excavation wall or a tunnel, the risk of a mountain slide, the risk of slip of a fault in the earth
crust or the force exerted on an oil platform in the Arctic by a moving mile-size ice ¯oe, the
scale jump from the laboratory spans many orders of magnitude.

Received 13 April 1999; accepted for publication 6 June 1999

Z. P. BazÏant
Walter P. Murphy Professor of Civil Engineering and Materials
Science, Northwestern University, Evanston, Illinois 60208, USA

Preparation of the present review article was supported by the

Of®ce of Naval Research under Grant N00014-91-J-1109 to
Northwestern University, monitored by Dr. Yapa D.S. Rajapakse.
 In most of mechanical and aerospace engineering, on the other hand, the problem of
scaling has been less pressing because the structures or structural components can usually
be tested at full size. It must be recognized, however, that even in that case the scaling
implied by the theory must be correct. Scaling is the most fundamental characteristic of any
physical theory. If the scaling properties of a theory are incorrect, the theory itself is
incorrect.
The size effect in solid mechanics is understood as the effect of the characteristic structure
size (dimension) D on the nominal strength rN of structure when geometrically similar
structures are compared. The nominal stress (or strength, in case of maximum load) is de®ned
as
704
cN P cN P
rN ˆ or
bD D2

for two- or three-dimensional similarity, respectively, where P is the load (or load
parameter), b structure thickness and cN arbitrary coef®cient chosen for convenience
(normally, cN ˆ 1). So rN is not a real stress but a load parameter having the dimension of
stress. The de®nition of D can be arbitrary (e.g. the beam depth or half-depth, the beam
span, the diagonal dimension, etc.) because it does not matter for comparing geometrically
similar structures.
The basic scaling laws in physics are power laws in terms of D, for which no characteristic
size (or length) exists. The classical Weibull [107] theory of statistical size effect caused by
randomness of material strength is of this type. During the 1970's it was found that a major
deterministic size effect, overwhelming the statistical size effect, can be caused by stress re-
distributions caused by stable propagation of fracture or damage and the inherent energy
release. The law of the deterministic size effect provides a way of bridging two different power
laws applicable in two adjacent size ranges. The structure size at which this bridging transition
occurs represents a characteristic size.
The material for which this new kind of size effect was identi®ed ®rst, and studied in the
greatest depth and with the largest experimental effort by far, is concrete. In general, a size
effect that bridges the small-scale power law for nonbrittle (plastic, ductile) behavior and the
large-scale power law for brittle behavior signals the presence of a certain nonnegligible
characteristic length of the material. This length, which represents the quintessential property
of quasibrittle materials, characterizes the typical size of material inhomogeneities or the
fracture process zone (FPZ). Aside from concrete, other quasibrittle materials include rocks,
cement mortars, ice (especially sea ice), consolidated snow, tough ®ber composites and
particulate composites, toughened ceramics, ®ber-reinforced concretes, dental cements, bone
and cartilage, biological shells, stiff clays, cemented sands, grouted soils, coal, paper, wood,
wood particle board, various refractories and ®lled elastomers as well as some special tough
metal alloys. Keen interest in the size effect and scaling is now emerging for various `high-tech'
applications of these materials.
Quasibrittle behavior can be attained by creating or enhancing material inhomogeneities.
Such behavior is desirable because it endows the structure made from a material incapable of
plastic yielding with a signi®cant energy absorption capability. Long ago, civil engineers
subconsciously but cleverly engineered concrete structures to achieve and enhance quasibrittle
characteristics. Most modern `high-tech' materials achieve quasibrittle characteristics in much
the same way ± by means of inclusions, embedded reinforcement, and intentional micro-
cracking (as in transformation toughening of ceramics, analogous to shrinkage microcracking
of concrete). In effect, they emulate concrete.
In materials science, an inverse size effect spanning several orders of magnitude must be
tackled in passing from normal laboratory tests of material strength to microelectronic com-
ponents and micromechanisms. A material that follows linear elastic fracture mechanics
(LEFM) on the scale of laboratory specimens of sizes from 1 to 10 cm may exhibit quasibrittle
or even ductile (plastic) failure on the scale of 0.1 or 100 microns.
The purpose of this article is to present a brief review of the basic results and their
history. For an in-depth review with a thousand of literature references, the recent article
[16] may be consulted. A full exposition of most of the material reviewed here is found in
the recent book [28]. The problem of scale bridging in the micromechanics of materials,
e.g. the relation of dislocation theory to continuum plasticity, is beyond the scope of this
review.
2
History of size effect up to Weibull
Speculations about the size effect can be traced back to Leonardo da Vinci (1500's) [112]. He
observed that ``among cords of equal thickness the longest is the least strong,'' and proposed
that ``a cord is so much stronger . . . as it is shorter,'' implying inverse proportionality. A
century later, Galileo Galilei (1638) [58], the inventor of the concept of stress, argued that
Leonardo's size effect cannot be true. He further discussed the effect of the size of an animal on
the shape of its bones, observing that bulkiness of bones is the weakness of the giants.
A major idea was spawned by Mariotte (1686) [76]. Based on his extensive experiments, he
observed that ``a long rope and a short one always support the same weight unless that in a long
rope there may happen to be some faulty place in which it will break sooner than in a shorter'', 705
and proposed the principle of ``the inequality of matter whose absolute resistance is less in one
place than another.'' In other words, the larger the structure, the greater is the probability of
encountering in it an element of low strength. This is the basic idea of the statistical theory of
size effect.
Despite no lack of attention, not much progress was achieved for two and half centuries,
until the remarkable work of Grif®th (1921) [60], the founder of fracture mechanics. He showed
experimentally that the nominal strength of glass ®bers was raised from 42,300 psi to
491,000 psi when the diameter decreased from 0.0042 in. to 0.00013 in., and concluded that
``the weakness of isotropic solids... is due to the presence of discontinuities or ¯aws... The
effective strength of technical materials could be increased 10 or 20 times at least if these ¯aws
could be eliminated.'' In Grif®th's view, however, the ¯aws or cracks at the moment of failure
were still only microscopic; their random distribution controlled the macroscopic strength of
the material but did not invalidate the concept of strength. Thus, Grif®th discovered the
physical basis of Mariotte's statistical idea but not a new kind of size effect.
The statistical theory of size effect began to emerge after Peirce (1926) [86], formulated the
weakest-link model for a chain and introduced the extreme value statistics which was origi-
nated by Tippett (1925) [101], Fischer and Tippett (1928) [52], and FreÂchet (1927) [51], and
re®ned by von Mises (1936) [102] and others (see also [50, 56, 57, 97]). The capstone of the
statistical theory was laid by Weibull (1939) [107] (also [108±110]). On a heuristic and ex-
perimental basis, he concluded that the tail distribution of low strength values with an ex-
tremely small probability could not be adequately represented by any of the previously known
distributions. He introduced what came to be known as the Weibull distribution, which gives
the probability of a small material element as a power law of the strength difference from a
®nite or zero threshold. Others (e.g. [56, 97]) later offered a theoretical justi®cation by means of
a statistical distribution of microscopic ¯aws or microcracks. Re®nement of applications to
metals and ceramics (fatigue embrittlement, cleavage toughness of steels at low and brittle-
ductile transition temperatures, evaluation of scatter of fracture toughness data) has continued
until today (e.g. [32, 35, 50, 71]).
Most subsequent studies of the statistical theory of size effect dealt basically with re®nements
and applications of Weibull's theory to fatigue embrittled metals and to ceramics (e.g. [69, 70]).
Applications to concrete, where the size effect was of the greatest concern, have been studied in
[36, 37, 78, 79, 82, 115, 116, 117] and elsewhere.
Until about 1985, most mechanicians paid almost no attention to the possibility of a de-
terministic size effect. Whenever a size effect was detected in tests, it was automatically as-
sumed to be statistical, and thus its study was supposed to belong to statisticians rather than
mechanicians. The reason probably was that no size effect is exhibited by the classical con-
tinuum mechanics in which the failure criterion is written in terms of stresses and strains:
elasticity with strength limit, plasticity and viscoplasticity, as well as fracture mechanics of
bodies containing only microscopic cracks or ¯aws [10]. The size effect was not even men-
tioned by S.P. Timoshenko in 1953 in his monumental History of the Strength of Materials.
The attitude, however, changed drastically in the 1980's. In consequence of the well-funded
research in concrete structures for nuclear power plants, theories exhibiting a deterministic size
effect have been developed. We will discuss it later.

3
Power scaling and the case of no size effect
It is proper to explain ®rst the simple scaling applicable to all physical systems that involve no
characteristic length. Let us consider geometrically similar systems, for example the beams
shown in Fig. 1a, and seek to deduce the response Y (e.g. the maximum stress or the maximum
706

Fig. 1. a Geometrically similar structures

of different sizes b power scaling laws c size
effect law for quasibrittle failures bridging
the power law of plasticity (horizontal
asymptote) and the power law of LEFM
(inclined asymptote)

de¯ection) as a function of the characteristic size (dimension) D of the structure; Y ˆ Y0 f …D†

where u is the chosen unit of measurement (e.g. 1 m, 1 mm). We imagine three structure sizes
1, D, and D0 (Fig. 1a). If we take size 1 as the reference sizes. The responses for sizes D and D0
are Y ˆ f …D† and Y 0 ˆ f …D0 †. However, since there is no characteristic length, we can also take
size D as the reference size. Consequently, the equation
 0
f …D0 † D
ˆf …1†
f …D† D

must hold [10, 16] (for ¯uid mechanics, see [4, 96]). This is a functional equation for the
unknown scaling law f …D†. It has one and only one solution, namely the power law:
 s
D
f …D† ˆ : …2†
c1

where s ˆ const. and c1 is a constant which is always implied as a unit of length measurement.
Note that c1 cancels out of Eq. (1) when the power function (2) is substituted.
On the other hand, when for instance f …D† ˆ log…D=c1 †, Eq. (1) is not satis®ed and the unit
of measurement, c1 , does not cancel out. So, the logarithmic scaling could be possible only if
the system possessed a characteristic length related to c1 .
The power scaling must apply for every physical theory in which there is no characteristic
length. In solid mechanics such failure theories include elasticity with a strength limit,
elastoplasticity, viscoplasticity as well as LEFM (for which the FPZ is assumed shrunken into a
point).
To determine exponent s, the failure criterion of the material must be taken into account. For
elasticity with a strength limit (strength theory), or plasticity (or elasto-plasticity) with a yield
surface expressed in terms of stresses or strains, or both, one ®nds that s ˆ 0 when response Y
represents the stress or strain (for example the maximum stress, or the stress at certain
homologous points, or the nominal stress at failure) [10]. Thus, if there is no characteristic
dimension, all geometrically similar structures of different sizes must fail at the same nominal
stress. By convention, this came to be known as the case of no size effect.
In LEFM, on the other hand, s ˆ ÿ1=2, provided that geometrically similar structures with
geometrically similar cracks or notches are considered. This may be generally demonstrated
with the help of Rice's J-integral [10].
If log rN is plotted versus log D, the power law is a straight line, Fig. 1. For plasticity, or
elasticity with a strength limit, the exponent of the power law vanishes, i.e. the slope of this line
is zero. For LEFM, the slope is ÿ1=2. A recently emerged `hot' subject is the quasibrittle
materials and structures, for which the size effect bridges these two power laws.
4
Weibull statistical size effect
The classical theory of size effect has been statistical. Three-dimensional continuous general-
ization of the weakest link model for the failure of a chain of links of random strength, Fig. 2
left, leads to the distribution
 Z 
Pf …rN † ˆ 1 ÿ exp ÿ c‰r…x†ŠdV…x† ; …3†
V

which represents the failure probability of a structure that fails as soon as a macroscopic
fracture initiates from a microcrack (or some ¯aw) somewhere in the structure; r is the stress 707
tensor ®eld just before failure, x the coordinate vector, V the volume of structure and c…r† a
function giving the spatial concentration of failure probability of the material (=Vrÿ1  failure
probability of material representative volume Vr ) [56],
P
i P1 …ri †
c…r† 
V0

where ri are the principal stresses (i ˆ 1; 2; 3) and P1 …r† the failure probability (cumulative) of
the smallest possible test specimen of volume V0 (or representative volume of the material)
subjected to uniaxial tensile stress r,
 
r ÿ ru m
P1 …r† ˆ ; …4†
s0

[107], where m; s0 ; r1 are material constants (m denotes the Weibull modulus, usually between
5 and 50, s0 is a scale parameter, ru the strength threshold, which may usually be taken as zero)
and V0 is the reference volume understood as the volume of specimens on which c…r† was
measured. For specimens under uniform uniaxial stress (and ru ˆ 0), (3) and (4) lead to the
following simple expressions for the mean and the coef®cient of variation of nominal strength:
 m1  12
V0 ÿ1 C…1 ‡ 2mÿ1 †
rN ˆ s0 C…1 ‡ m † ; xˆ ÿ1 ; …5†
V C2 …1 ‡ mÿ1 †

where C is the gamma function. Since x depends only on m, it is often used for determining m
from the observed statistical scatter of strength of identical test specimens. The expression for
rN includes the effect of volume V which depends on size D. In general, for structures with
nonuniform multidimensional stress, the size effect of Weibull theory (for rr  0) is of the
type:

Fig. 2. a Chain with many links of

random strength b failure proba-
bility of a small element c struc-
ture with many microcracks of
different probabilities to become
critical
 nd
rN / Dÿ m ; …6†

where nd ˆ 1; 2 or 3 for uni-, two- or three-dimensional similarity.

In view of (5), the value rW ˆ rN …V=V0 †1=m for a uniformly stressed specimen can be
adopted as a size-independent stress measure. Taking this viewpoint, Beremin [32], proposed
taking into account the nonuniform stress in a large crack-tip plastic zone by the so-called
Weibull stress:
!m1
X Vi
rW ˆ rm
Ii ; …7†
708 i
V0

where Vi …i ˆ 1; 2; . . . ; NW † are elements of the plastic zone having maximum principal stress
rIi . The sum in (5) was replaced by an integral in [95], see also [71]. Equation (7), however,
considers only the crack-tip plastic zone whose size which is almost independent of D. Con-
sequently, Eq. (7) is applicable only if the crack at the moment of failure is not yet macroscopic,
still being negligible compared to structural dimensions.
As far as quasibrittle structures are concerned, applications of the classical Weibull theory
face a number of serious objections:
1. The fact that in (6) the size effect is a power law implies the absence of any characteristic
length. But this cannot be true if the material contains sizable inhomogeneities.
2. The energy release due to stress redistributions caused by macroscopic FPZ, or stable crack
growth before Pmax , gives rise to a deterministic size effect which is ignored. Thus the Weibull
theory is valid only if the structure fails as soon as a microscopic crack becomes macroscopic.
3. Every structure is mathematically equivalent to a uniaxially stressed bar (or chain, Fig. 2),
which means that no information on the structural geometry and failure mechanism is taken
into account.
4. The size-effect differences between two- and three-dimensional similarities (nd ˆ 2 or 3) are
predicted much too large.
5. Many tests of quasibrittle materials (e.g. diagonal shear failure of reinforced concrete beams)
show a much stronger size effect than predicted by Weibull theory, [28] and the review in [11]).
6. The classical theory neglects the spatial correlations of material failure probabilities of
neighboring elements caused by nonlocal properties of damage evolution (while general-
izations based on some phenomenological load-sharing hypotheses have been divorced from
mechanics).
7. When Eq. (5) is ®t to the test data on statistical scatter for specimens of one size (V =
const.), and when Eq. (6) is ®t to the mean test data on the effect of size (of unnotched plain
concrete specimens), the optimum values of Weibull exponent m are very different, namely
m ˆ 12 and m ˆ 24, respectively. If the theory were applicable, these value would have to
coincide.
In view of these limitations, among concrete structures Weibull theory appears applicable to
some extremely thick plain (unreinforced) structures, e.g. the ¯exure of an arch dam acting as a
horizontal beam (but not for vertical bending or arch dams nor gravity dams, because large
vertical compressive stresses cause long cracks to grow stably before the maximum load). Most
other plain concrete structures are not thick enough to prevent the deterministic size effect
from dominating. Steel or ®ber reinforcement prevents it as well.

5
Quasibrittle size effect bridging plasticity and LEFM, and its history
Quasibrittle materials are those that obey on a small scale the theory of plasticity (or strength
theory), characterized by material strength or yield limit r0 , and on a large scale the LEFM,
characterized by fracture energy Gf . While plasticity alone, as well as LEFM alone, possesses no
characteristic length, the combination of both, which must be considered for the bridging of
plasticity and LEFM, does. Combination of r0 and Gf yields Irwin's (1958) characteristic length
(material length)

EGf
`0 ˆ ; …8†
r20
which approximately characterizes the length of the FPZ; E is Young's elastic modulus. So
the key to the deterministic quasibrittle size effect is a combination of the concept of strength
or yield with fracture mechanics. In dynamics, this further implies the existence of a charac-
teristic time (material time):

s0 ˆ l0 =v ; …9†

representing the time a wave of velocity v takes to propagate the distance l0 .

After LEFM was ®rst applied to concrete [66], it was found to disagree with test re-
sults [68, 72, 105, 106]. In [72], Leicester tested geometrically similar notched beams of
different sizes, ®t the results by a power law, rN / Dÿn , and observed that the optimum n
709
was less than 1/2, the value required by LEFM. The power law with a reduced exponent, of
course, ®ts the test data in the central part of the transitional size range well but does
not provide the bridging of the ductile and LEFM size effects. It was tried to explain the
reduced exponent value by notches of a ®nite angle, which however is objectionable for two
reasons: (i) notches of a ®nite angle cannot propagate (rather, a crack must emanate from
the notch tip), (ii) the singular stress ®eld of ®nite-angle notches gives a zero ¯ux of energy
into the notch tip. Same as the Weibull theory, Leicester's power law also implied non-
existence of a characteristic length (see [16], Eqs. (1)±(3)); which cannot be the case for
concrete due to the large size of its inhomogeneities. More extensive tests of notched
geometrically similar concrete beams of different sizes were carried out by Walsh [105, 106].
Although he did not attempt a mathematical formulation, he was the ®rst to make the
doubly logarithmic plot of nominal strength versus size and observe that it was transitional
between plasticity and LEFM.
An important advance was made in [62] and also in [87]. Inspired by the softening and
plastic FPZ models of Barenblatt [2, 3] and Dugdale [49], Hillerborg et al. formulated the
cohesive (or ®ctitious) crack model characterized by a softening stress-displacement law for the
crack opening and showed by ®nite element calculations that the failures of unnotched plain
concrete beams in bending exhibit a deterministic size effect, in agreement with tests of the
modulus of rupture.
Analyzing distributed (smeared) cracking damage, BazÏant [5] demonstrated that its local-
ization into a crack band engenders a deterministic size effect on the postpeak de¯ections and
energy dissipation of structures. The effect of the crack band is approximately equivalent to
that of a long fracture with a sizable FPZ at the tip. Subsequently, using an approximate energy
release analysis, the following approximate size effect law was derived in [6] for the quasibrittle
size effect in structures failing after large stable crack growth:

  1
D ÿ2
rN ˆ Br0 1 ‡ ‡ rR ; …10†
D0

or more generally:
  r ÿ2r1
D
rN ˆ Br0 1 ‡ ‡ rR ; …11†
D0

in which r; B are positive dimensionless constants, D0 = const. represents the transitional

size at which the power laws of plasticity and LEFM intersect; D0 and B characterize the
structure geometry. Usually, constant rR ˆ 0, except when there is a residual crack-bridging
stress rr outside the FPZ, as in ®ber composites, or when at large sizes some plastic
mechanism acting in parallel emerges and becomes dominant, as in the Brazilian split-
cylinder test.
Equation (10) was shown to be closely followed by the numerical results for the crack band
model [5, 26], as well as for the nonlocal continuum damage models, which are capable of
realistically simulating the localization of strain-softening damage and avoiding spurious mesh
sensitivity.
Beginning in the mid 1980's, the interest in the quasibrittle size effect of concrete structures
surged enormously and many researchers made noteworthy contributions; to name but a few:
[36, 87, 88±90]. The size effect has recently become a major theme at conferences on concrete
fracture [8, 30, 80, 81, 114].
 Measurements of the size effect of Pmax were shown to offer a simple and reliable way to
determine the fracture characteristics of quasibrittle materials, including the fracture energy,
effective FPZ length, the critical crack tip opening displacement, and geometry-dependent R-
curve.

6
Size-effect mechanism: stress redistribution and energy release
Let us now describe the gist of the deterministic quasibrittle size effect. LEFM applies when the
FPZ is negligibly small compared to structural dimension D and can be considered as a point.
Thus the LEFM solutions can be obtained by methods of elasticity. The salient characteristic of
710 quasibrittle materials is that there exists a sizable FPZ with distributed cracking or other
softening damage that is not negligibly small compared to structural dimension D. This makes
the problem nonlinear, although approximately equivalent LEFM solutions can be applied,
unless FPZ reaches near the structure boundaries.
The existence of a large FPZ means that the distance between the tip of the actual traction-
free crack and the tip of the equivalent LEFM crack at Pmax is equal to a certain characteristic
length cf (roughly one half of the FPZ size) that is not negligible compared to D. This causes a
nonnegligible macroscopic stress redistribution with energy release from the structure.
With respect to the fracture length a0 (distance from the mouth of notch or crack to the
beginning of the FPZ), two basic cases may now be distinguished: (i) a0 ˆ 0, which means that
Pmax occurs at the initiation of macroscopic fracture propagation, and (ii) a0 is ®nite and not
negligible compared to D, which means that Pmax occurs after large stable fracture growth.

6.1
Scaling for failure at crack initiation
An example of the ®rst case is the modulus of rupture test, which consists in the bending of a
simply supported beam of span L with a rectangular cross section of depth D and width b,
subjected to concentrated load P, the maximum load is not decided by the stress
r1 ˆ 3PL=2bD2 at the tensile face, but by the stress value r  roughly at distance cf =2 from the
tensile face, which is at the middle of FPZ. Because r  ˆ r1 ÿ r01 cf =2, where r01 is the stress
 ˆ r0 is the intrinsic tensile strength of the material,
gradient equal to 2r1 =D, and also because r
the failure condition r ˆ r0 yields P=bD ˆ rN ˆ r0 =…1 ÿ Db =D†, where Db ˆ …3L=2D†cf , which
is a constant because for geometrically similar beams L=D = const. This expression, however,
is unacceptable for D  Db . But since the derivation is valid only for small enough cf =D, one
may replace it by the following asymptotically equivalent size effect formula:
 1r
rDb
r N ˆ r0 1‡ ; …12†
D

which happens to be acceptable for the whole range of D (D < material inhomogeneity size is
excluded, of course); r is any positive constant. The values r ˆ 1 or 2 have been used for
concrete [7], while r  1:45 is optimum according to the latest analysis of test data at
Northwestern University (yet unpublished).

6.2
Scaling for failures with a long crack or notch
Let us now give a simple explanation of the second case of structures failing only after stable
formation of large cracks, or notched fracture specimens. Failures of this type, exhibiting a
strong size effect [18, 28, 59, 63, 77, 98, 104] are typical of reinforced concrete structures or ®ber
composites [17, 113], and are also exhibited by some unreinforced structures (e.g. dams, due to
the effect of vertical compression, or ¯oating ice plates in the Arctic). Consider the rectangular
panel in Fig. 3, which is initially under a uniform stress equal to rN . Introduction of a crack of
length a with a FPZ of a certain length and width h may be approximately imagined to relieve
the stress, and thus release the strain energy, from the shaded triangles on the ¯anks of the
crack band shown in Fig. 3. The slope k of the effective boundary of the stress-relief zone need
not be determined; what is important is that k is independent of the size D.
For the usual ranges of interest, the length of the crack at maximum load may normally be
assumed approximately proportional to the structure size D, while the size h of the FPZ is
essentially a constant, related to the inhomogeneity size in the material. This has been veri®ed
 Fig. 3. Approximate zones of stress re-
lief due to fracture 711

for many cases by experiments (showing similar failure modes for small and large specimens)
and ®nite element solutions based on the crack band, cohesive or nonlocal models.
The stress reduction in the triangular zones of areas ka2 =2, Fig. 3, causes (for the case b ˆ 1)
the energy release Ua ˆ 2  …ka2 =2†r2N =2E. The stress drop within the crack band of width h
causes further energy release Ub ˆ har2N =E. The total energy dissipated by the fracture is
W ˆ aGf , where Gf is the fracture energy, a material property representing the energy dissi-
pated per unit area of the fracture surface. Energy balance during static failure requires that
o…Ua ‡ Ub †=oa ˆ dW=da. Setting a ˆ D…a=D†, where a=D is approximately a constant if the
failures for different structure sizes are geometrically similar, the solution of the last equation
for rN yields [6] the approximate size effect law in (10) with rR ˆ 0, Fig. 1 bottom.
More rigorous derivations of this law, applicable to arbitrary structure geometry, have been
given in terms of asymptotic analysis based on equivalent LEFM [12] or on Rice's path-
independent J-integral [28]. This law has also been veri®ed by nonlocal ®nite element analysis,
by the cohesive crack model, and by random particle (or discrete element) models. The
experimental veri®cations, among which the earliest is found in the famous Walsh's tests
[105, 106] of notched concrete beams, have by now become abundant, e.g. Fig. 4.
For very large sizes …D  D0 †, the size effect law in (10) reduces to the power law rN / Dÿ1=2 ,
which represents the size effect of LEFM for geometrically similar large cracks and corresponds
to the inclined asymptote of slope ÿ1=2 in Fig. 1 (bottom). For very small sizes …D  D0 †, this
law reduces to rN = const.; which corresponds to the horizontal asymptote and means that
there is no size effect, as in plastic limit analysis.
The ratio b ˆ D=D0 is called the brittleness number of a structure. For b ! 1, the structure
is perfectly brittle, i.e. follows LEFM, in which case the size effect is the strongest possible, while
for b ! 0 the structure is nonbrittle (or ductile, plastic), in which case there is no size effect.
Quasibrittle structures are those for which 0:1  b  10, in which case the size effect represents
a smooth transition (or interpolation) that bridges the power law size effects for the two
asymptotic cases. The law (10) has the character of asymptotic matching and serves to provide
the bridging of scales. In the quasibrittle range, the stress analysis is of course nonlinear, calling
for the cohesive crack model or the crack band model (which are mutually almost equivalent),
or some of the nonlocal damage models.
The meaning of the term quasibrittle is relative. If the size of a quasibrittle structure becomes
suf®ciently large compared to material inhomogeneities, the structure becomes perfectly brittle
(for concrete structures, only the global fracture of a large dam is describable by LEFM), and if the
size becomes suf®ciently small, the structure becomes nonbrittle (plastic, ductile), because the
FPZ extends over the whole cross section of the structure. Thus, a micromachine or a miniature
electronic device made of silicone or ®ne-grained ceramic may be quasibrittle or nonbrittle.

6.3
Size effect on postpeak softening and ductility
The plots of nominal stress versus the relative structure de¯ection (normalized so as to make
the initial slope size-independent) have, for small and large structures, the shapes indicated in
Fig. 5. Apart from the size effect on Pmax , there is also a size effect on the shape of the postpeak
descending load-de¯ection curve. For small structures the postpeak curves descend slowly, for
larger structures steeper, and for suf®ciently large structures they may exhibit a snapback, that
is, a change of slope from negative to positive.
If a structure is loaded under displacement control through an elastic device with spring
constant Cs , it loses stability and fails at the point where the load-de¯ection diagram ®rst attains
712

Fig. 4. a Comparisons of size effect law with Mode I test data obtained by various investigators using
notched specimens of different materials b size effect compression kink-band failures of geometrically
similar notched carbon-PEEK specimens (after [20])

the slope ÿCs (if ever), Fig. 5. The ratio of the de¯ection at these points to the elastic de¯ection
characterizes the ductility of the structure. As apparent from the ®gure, small quasibrittle
structures have a large ductility while large quasibrittle structures have small ductility.
The areas under the load-de¯ection curves in Fig. 5 characterize the energy absorption. The
capability of a quasibrittle structure to absorb energy decreases, in relative terms, as the
structure size increases. The size effect on energy absorption capability is important for blast
loads and impact.
 713

Fig. 5. Load-de¯ection curves of quasibrittle

structures of different sizes, scaled to the same
initial slope

The progressive steepening of the postpeak curves in Fig. 5 with increasing size and the
development of a snapback can be most simply described by the series coupling model,
which assumes that the response of a structure may be at least approximately modeled by
the series coupling of the cohesive crack or damage zone with a spring characterizing the
elastic unloading of the rest of the structure ([15], Sec. 13.2).

6.4
Asymptotic analysis of size effect by equivalent LEFM
To obtain simple approximate size-effect formulae that give a complete prediction of the
failure load, including the effect of geometrical shape of the structure, equivalent LEFM may
be used. In this approach the tip of the equivalent LEFM sharp crack is assumed to lie
approximately a distance cf ahead of the tip of the traction-free crack or notch, cf being a
constant representing roughly one half of the length of the FPZ ahead of the tip. Two cases
are relatively simple:
(i) If a large crack grows stably prior to Pmax or if there is a long notch
p p
EGf ‡ rY c0 …a0 †cf ‡ c…a0 †D
rN ˆ p ; …13†
g 0 …a0 †cf ‡ g…a0 †D

(ii) If Pmax occurs at fracture initiation from a smooth surface

q2
p c
EGf ‡ rY c0 …0†cf ‡ c00 …0† 2Df
rN ˆ q2 ; …14†
cf
0
g …0†cf ‡ g …0† 2D 00

2
[7, 12], where the primes denote derivatives; g…a0 † ˆ KIP =r2N D and c…a0 † ˆ KIr
2
=r2Y D are
dimensionless energy-release functions of LEFM of a ˆ a0 =D, where a0 is the length of
notch or crack up to the beginning of the FPZ; KIP ; KIr are stress intensity factors for load
P and for loading by uniform residual crack-bridging stress rY , respectively; rY > 0 for
tensile fracture, but rY 6ˆ 0 in the cases of compression fracture in concrete, kink band
propagation in ®ber composites, and tensile fracture of composites reinforced by ®bers
short enough to undergo frictional pullout rather than breakage. The asymptotic behavior
of (13) for D ! 1 is of the LEFM type, rN ÿ rY / Dÿ1=2 . Formula (14) approaches for
D ! 1 a ®nite asymptotic value; so does formula (13), if rY > 0.
 6.5
Size-effect method for measuring material constants and the R-curve
Comparison of (13) with (10) yields the relations
s
g 0 …a0 † EGf
D0 ˆ cf ; Br0 ˆ r0 : …15†
g…a0 † cf g 0 …a0 †

Therefore, by ®tting formula (10) with rR ˆ 0 to the values of rN measured on test

specimens of different sizes with a suf®ciently broad range of brittleness numbers
714 b ˆ D=D0 , the values of Gf and cf can be identi®ed [19, 28]. The ®tting can best be done by
using the Levenberg±Marquardt nonlinear optimization algorithm, but it can also be ac-
complished by a properly weighted linear regression of rÿ2 N versus D. The specimens do not
have to be geometrically similar, although, when they are, the evaluation is simpler and the
error smaller. The lower the scatter of test results, the narrower is the minimum necessary
range of b (for concrete and ®ber composites, the size range 1:4 is the minimum).
The size-effect method of measuring fracture characteristics has been adopted for an in-
ternational standard recommendation for concrete ([28, 93] Sec. 6.3), and has also been veri®ed
and used for various rocks, ceramics, orthotropic ®ber-polymer composites, sea ice, wood,
tough metals and other quasibrittle materials. The advantage of the size-effect method is that
the tests, requiring only the maximum loads, are foolproof and easy to carry out. With regard
to the cohesive crack model, note that the size-effect method gives the energy value corre-
sponding to the area under the initial tangent of the softening stress-displacement curve, rather
than the total area under the curve.
The size-effect method also permits determining the R-curve (resistance curve) of the
quasibrittle material, a curve that represents the apparent variation of fracture energy with
crack extension for which LEFM becomes approximately equivalent to the actual material
with a large FPZ. The R-curve, which depends on the specimen geometry, in contrast to the
classical R-curve de®nition, can be obtained as the envelope of the curves of the energy
release rate at P ˆ Pmax (for each size) versus the crack extension for specimens of various
sizes (or various brittleness numbers). In general, this can easily be done numerically, and
if the size-effect law has the form in (10) with rR ˆ 0, a parametric analytical expression for
the R-curve exists ([19, 28], Sec. 6.4).
The fracture model implied by the size-effect law in (10) or (13) has one independent
characteristic length, cf , representing about one half of the FPZ length. Because of (15), the
value of `0 is implied by cf if r0 is known. The value of cf controls the size D0 at the center of
the bridging region, intersection of the power-law asymptotes in Fig. 1 (bottom), and r0 or Gf
control a vertical shift of the size-effect curve at constant D0 . The location of the large-size
asymptote depends only on Kc and geometry, and the location of the small-size asymptote
depends only on r0 and geometry.

6.6
Critical crack-tip opening displacement, dCTOD
The quasibrittle size effect, bridging plasticity and LEFM, can also be simulated by the fracture
models characterized by the critical stress intensity factor Kc (fracture toughness) and dCTOD ;
for metals see [45, 111], and for concrete [64]. Jenq and Shah's model [64], called the two-
parameter fracture model, has been shown to give essentially the same results as the R-curve
derived from the size-effect law in (10) with rR ˆ 0 . The models are in practice equivalent
because
r
p 1 BGf cf
Kc ˆ EGf ; dCTOD ˆ …16†
p E

[18]. Using these formulae, the values of Kc and dCTOD can be easily identi®ed by ®tting the size
effect law (10) to measured Pmax value.
Like the size-effect law in (10) with rR ˆ 0, the two-parameter model has only one inde-
pendent characteristic length, `0 ˆ Kc2 =r20 . If r0 is known, then dCTOD is not an independent
length because cf is implied by `0 , and dCTOD then follows from (16).
7
Extensions, ramifications and applications

7.1
Size effects in compression fracture
Loading by high compressive stress without suf®cient lateral con®ning stresses leads to damage
in the form axial splitting microcracks engendered by pores, inclusions or inclined slip planes.
This damage localizes into a band that propagates either axially or laterally.
For axial propagation, the energy release from the band drives the formation of the axial
splitting fracture, and since this energy release is proportional to the length of the band, there is
no size effect. For lateral propagation, the stress in the zones on the sides of the damage band 715
gets reduced, which causes an energy release that grows in proportion to D2 , while the energy
consumed and dissipated in the band grows in proportion to D. The mismatch of energy-
release rates inevitably engenders a deterministic size effect of the quasibrittle type, analogous
to the size effect associated with tensile fracture. In consequence of the size effect, failure by
lateral propagation must prevail over the failure by axial propagation if a certain critical size is
exceeded.
The size effect can again be approximately described by the equivalent LEFM. This leads to
Eq. (13), in which rY is determined by analysis of the microbuckling in the laterally propa-
gating band of axial splitting cracks. The spacing s of these cracks is in (13) assumed to be
dictated by material inhomogeneities. However, if the spacing is not dictated, and is such that it
minimizes rN , then the size effect gets modi®ed as
2
rN ˆ CDÿ5 ‡ r1 ; …17†

([28], Sec. 10.5.11) where C; r1 are constants, the approximate values of which have been
calculated for the breakout of boreholes in rock.

7.2
Fracturing truss model for concrete and boreholes in rock
Propagation of compression fracture is what appears to control the maximum load in diagonal
shear failure of reinforced concrete beams with or without stirrups, for which a very strong
size effect has been demonstrated experimentally, e.g. [11, 19, 63, 65, 85, 92, 98, 103±104]. A
long diagonal tension crack grows stably under shear loading until the concrete near its tip gets
crushed. A simpli®ed formula for the size effect can be obtained by energetic modi®cation of
the classical truss model (strut-and-tie model) [11].
The explosive breakout of boreholes (or mining stopes) in rock under very high pressures is
known to also exhibit size effect, as revealed by the tests in [42, 43, 61, 84]. An approximate
analytical solution can be obtained by exploiting Eshelby's theorem for eigenstresses in
elliptical inclusions [25].

7.3
Kink bands in fiber composites
A kink band, in which axial shear-splitting cracks develop between ®bers which undergo
microbuckling, is one typical mode of compression failure of composites or laminates with
uniaxial ®ber reinforcement. This failure mode, whose theory was begun in [1, 94], was until
recently treated by the theory of plasticity, which implies no size effect. Recent experimental
and theoretical studies [35], however, revealed that the kink band propagates side-way like a
crack, and the stress on the ¯anks of the band gets reduced to a certain residual value, which is
here denoted as rY and can be estimated by the classical plasticity approach of Budianski [34].
The crack-like behavior implies a size effect, which is demonstrated by the latest laboratory
tests of notched carbon-PEEK specimens [20], Fig. 4; these tests also demonstrated the pos-
sibility of a stable growth of a long kink band, which was achieved by rotational restraint at the
ends.
There are again two types of size effect, depending on whether Pmax is reached (i) when the
FPZ of the kink band is attached to a smooth surface or (ii) or when there exists either a notch
or a long segment of the kink band in which the stress has been reduced to rY . Formulae (13)
and (14), respectively, approximately describe the size effects for these two basic cases; in this
case, Gf plays the role of fracture energy of the kink band (area below the stress-contraction
 curve of the kink band and above the rY value), and cf -the role of the FPZ of the kink band,
which is assumed to be approximately constant, governed by material properties.
The aforementioned carbon-PEEK tests also con®rm that case (ii), in which a long kink band
grows stably prior to Pmax , is possible; in those tests, stability is achieved by virtue of a lateral
shift of compression resultant in wide notched prismatic specimens with ends restrained
against rotation.

7.4
Size effects in sea ice
Normal laboratory specimens of sea ice exhibit no notch sensitivity. Therefore, failure of sea ice
716 has been thought to be well-described by plastic limit analysis, which exhibits no size effect, e.g.
[66, 100]. This perception, however, changed drastically after Dempsey carried out in 1993 on
the Arctic Ocean size-effect tests of ¯oating notched square specimens with an unprecedented,
record-breaking size range (with square sides ranging from 0.5 m to 80 m!), see [46, 47, 83].
It is now clear that ¯oating sea ice plates are quasibrittle and their size effect on the scale of
100 m approaches that of LEFM. Among other things, Dempsey's major experimental result
(Fig. 4 middle) explains why the measured forces exerted by moving ice on a ®xed oil platform
are one to two orders of magnitude smaller than the predictions of plastic limit analysis based on
the laboratory strength of ice. The size effect law in (10) with rR ˆ 0, or in (13) with rY ˆ 0,
agrees with these results well, permitting the values of Gf and cf of sea ice to be extracted by
linear regression of the Pmax data. The value of cf is in the order of meters, which can be
explained by inhomogeneities such as brine pockets and channels, as well as preexisting thermal
cracks, bottom roughness of the plate, warm and cold spots due to alternating snow drifts, etc..
Information on the size effect in sea ice can also be extracted from acoustic measurements [74].
Rapid cooling in the Arctic can in the ¯oating plate produce bending moments large enough
to cause fracture. According to plasticity or elasticity with a strength limit, the critical tem-
perature difference DTcr across the plate would have to be independent of plate thickness D.
Fracture analysis, however, indicated a quasibrittle size effect. Curiously, its asymptotic form is
not DTcr / Dÿ1=2 but
3
DTcr / Dÿ8 ; …18†

[9]. The reason is that D is not a characteristic dimension in the plane of the boundary value
problem of plate bending; rather it is the ¯exural wavelength of a plate on elastic foundation,
which is proportional to D4=3 rather than D. It seems that (18) may explain why long cracks of
length 10 to 100 km, which suddenly form in the fall in the Arctic ice cover, often run through
thick ice ¯oes and do not follow the thinly refrozen water leads around the ¯oes.
In analyzing the vertical penetration of ¯oating ice plate (load capacity for heavy objects on
ice, or the maximum force P required for penetration from below), one must take into account
that bending cracks reach only through part of the thickness, their ligaments transmitting
compressive forces which produces a dome effect. Because at maximum load the part-through
bending crack (of variable depth pro®le) is growing vertically, the asymptotic size effect is not
P=D2 ˆ rN / Dÿ3=8 [99], but rn / Dÿ1=2 . This was determined by a simpli®ed analytical so-
lution (with a uniform crack depth) in [48], and con®rmed by a detailed numerical solution
with a variable crack depth pro®le [21]. The latter also led to an approximate prediction
formula for the entire practical range of D, which is of the type of (10) with rN ˆ 0. This
formula was shown to agree with the existing ®eld tests [53, 54, 75].

7.5
Influence of crack separation rate, creep and viscosity
There are two mechanisms in which the loading rate affects fracture growth: (i) creep of the
material outside the FPZ, and (ii) rate dependence of the severance of material bonds in the
FPZ. The latter may be modeled as a rate process controlled by activation energy, with Ar-
rhenius-type temperature dependence. This leads to a dependence of the softening stress-
separation relation of the cohesive crack model on the rate of opening displacement. In an
equivalent LEFM approach, the latter is modeled by considering the crack-extension rate to be
a power function of the ratio of the stress intensity factor to its critical R-curve value.
For quasibrittle materials exhibiting creep (e.g. concretes and polymer composites, but not
rocks or ceramics), the consequence of mechanism 1 (creep) is that a decrease of loading rate,
or an increase of duration of a sustained load, causes a decrease of the effective length of the
FPZ. This in turn means an increase of the brittleness number manifested by a leftward rigid-
body shift of the size-effect curve in the plot of log rN versus log D, i.e. a decrease of effective
D0 . For slow or long-time loading, quasibrittle structures become more brittle and exhibit a
stronger size effect [24].
Mechanism 2 (rate-dependence of separation) causes that an increase of loading rate, or a
decrease of sustained load duration, leads to an upward vertical shift of the size effect curve for
log rN but has no effect on D0 and thus on brittleness. This mechanism also explains an
interesting recently discovered phenomenon ± a reversal of softening to hardening after a
sudden increase of the loading rate, which cannot be explained by creep.
So far all our discussions dealt with statics. In dynamic problems, any type of viscosity g of
717
the material (present in models for creep, viscoelasticity or viscoplasticity) implies a charac-
teristic length. Indeed, since g means stress/strain rate = kg/ms, and the Young's modulus E
and mass density q have dimensions E ˆ kg/ms2 and q ˆ kg/m3 , the material length associated
with viscosity is given by
s
g E
`v ˆ ; vˆ ; …19†
vq q
where v is the wave velocity. Consequently, any rate-dependence in the constitutive law implies
a size effect (and a nonlocal behavior as well).
There is, however, an important difference. Unlike the size effect associated with `0 or cf , the
viscosity-induced size effect (as well as the width of damage localization zones) is not time-
independent. It varies with the rates of loading and deformation of the structure, and vanishes
as the rates drop to zero. For this reason, an arti®cial viscosity or rate effect can approximate
the nonviscous size effect and localization only within a narrow range of time delays and rates,
but not generally.
7.6
Size effect in fatigue crack growth
Cracks slowly grow under fatigue (repeated) loading. This is for metals and ceramics described
by the Paris (or Paris-Erdogan) law, which states that the plot of the logarithm of the crack
length increment per cycle versus the amplitude of the stress intensity factor is a rising straight
line. For quasibrittle material it turns out that a size increase causes this straight line to shift to
the right, the shift being derivable from the size effect law in (10) ([28], Sec. 11.7).
7.7
Size effect for cohesive crack model and crack band model
The cohesive (or ®ctitious) crack model (called by Hillerborg et al., [62], and Petersson, [87],
the ®ctitious crack model) is more accurate yet less simple than the equivalent LEFM. It is based
on the hypothesis that the three-dimensional behavior of a ®nite FPZ can be approximated by
a unique one-dimensional decreasing function w ˆ g…r† relating the crack-opening displace-
ment w (separation of crack faces) to the crack bridging stress r in the FPZ. The obvious way to
determine the size effect is to solve Pmax by numerical integration for step-by-step loading [87].
The size-effect plot, however, can be solved directly if one inverts the problem, searching
the size D for which a given relative crack length a ˆ a=D corresponds to Pmax . This leads to
the equations [73]:
Z a Ra
v…n†dn
D C …n; n †v…n †dn ˆ ÿg ‰r…n†Šv…n†; Pmax ˆ R a a0rP
rr 0 0 0 0
; …20†
a0 D a0 C …n†v…n†dn

where the ®rst represents an eigenvalue problem for a homogeneous Fredholm integral
equation, with D as the eigenvalue and v…n† as the eigenfunction; n ˆ x=D; x is the coordinate
along the crack, Fig. 6; a ˆ a=D; a0 ˆ a0 =D; a; a0 are the total and the traction-free crack
lengths (or notch lengths); and C rr …n; n0 †; C rP …n† denote the compliance functions of structure
for crack surface force and given load P. Choosing a sequence of a-values, one obtains for each,
from (20), the corresponding values of D and Pmax . These results have also been generalised to
obtain directly the load and displacement corresponding, on the load-de¯ection curve, to a
point with any given tangential stiffness, including the displacement at the snapback point
which characterizes the ductility of the structure.
718 Fig. 6. Cohesive crack and distribution of crack-
bridging stresses

The cohesive crack model possesses at least one, but for concrete typically two, independent
characteristic lengths: `0 ˆ EGf =r20 and `1 ˆ EGF =r20 , where GF is the area under the entire
softening stress±displacement curve r ˆ f …w†, and Gf is the area under the initial tangent to
this curve, which is equal to GF only if the curve is simpli®ed as linear (typically GF  2Gf ). The
bilinear stress-displacement law used for concrete involves further parameters of the length
dimension ± the opening displacement wf when the stress is reduced to zero, and the dis-
placement at the change of slope; but their values are implied by Gf ; GF ; r0 and the stress at the
change of slope. The minimum admissible spacing h of parallel cohesive cracks is also a
necessary parameter.
The scatter of size-effect measurements within a practicable size range (up to 1:30) normally
does not permit identifying more than one characteristic length (measurements of postpeak
behavior are used for that purpose). Vice versa, when only the maximum loads of structures in
the bridging region between plasticity and LEFM are of interest, hardly more than one char-
acteristic length (namely cf ) is needed.
The crack band model, which is easier to implement and is used in commercial codes
(e.g. DIANA, SBETA [44]), is, for localized cracking or fracture, nearly equivalent to the
cohesive crack model [28, 91], provided that the effective (average) transverse strain in the
crack band is taken as ey ˆ w=h, where h is the width of the band. All that has been said
about the cohesive crack model also applies to the crack band model. Width h, of course,
represents an additional characteristic length, `4 ˆ h. It matters only when the cracking is
not localized but distributed (e.g. due to the effect of dense and strong enough rein-
forcement), and it governs the minimum spacings of parallel cracks. Their spacing cannot
be unambiguously captured by the standard cohesive crack model.
7.8
Size effect via nonlocal, gradient or discrete element models
The hypostatic feature of any model capable of bridging the power law size effects of
plasticity and LEFM is the presence of some characteristic length, `. In the equivalent LEFM
associated with the size effect law in (10), cf serves as a characteristic length of the ma-
terial, although this length can equivalently be identi®ed with dCTOD in Wells-Cottrell or
Jenq-Shah models, or with the crack opening wf at which the stress in the cohesive crack
model (or crack band model) is reduced to zero (for size effect analysis with the cohesive
crack model, see [23, 28]).
In the integral-type nonlocal continuum damage models, ` represents the effective size of the
representative volume of the material, which in turn plays the role of the effective size of the
averaging domain in nonlocal material models. In the second-gradient nonlocal damage
models, which may be derived as an approximation of the nonlocal damage models, a material
length is involved in the relation of the strain to its Laplacian. In damage simulation by the
discrete element (or random particle) models, the material length is represented by the sta-
tistical average of particle size.
The existence of ` in these models engenders a quasibrittle size effect that bridges the power-
law size effects of plasticity and LEFM and follows closely Eq. (10) with rN ˆ 0, as documented
by numerous ®nite element simulations. It also poses a lower bound on the energy dissipation
during failure, prevents spurious excessive localization of softening continuum damage, and
eliminates spurious mesh sensitivity ([28], Ch. 13).
These important subjects will not be discussed here any further because there exists a recent
extensive review [14].
7.9
Nonlocal statistical generalization of Weibull theory
Two cases need to be distinguished: (a) the front of the fracture that causes failure can lie at
only one place in the structure, or (b) the front can lie, with different probabilities, at many
different places. The former case occurs when a long crack, whose path is dictated by fracture
mechanics, grows before attaining the maximum load, or if a notch is cut in a test specimen.
The latter case occurs when the maximum load is achieved at the initiation of fracture growth.
In both cases, the existence of a large FPZ calls for a modi®cation of Weibull concept: the
failure probability P1 at a given point of the continuous structure depends not on the local
stress at that point, but on the nonlocal strain, which is calculated as the average of the local
strains within the neighborhood of the point constituting the representative volume of the 719
material. The nonlocal approach broadens the applicability of Weibull concept to the case of
notches or long cracks, for which the existence of crack-tip singularity causes the classical
Weibull probability integral to diverge at realistic m-values. In cleavage fracture of metals, the
problem of crack singularity has been circumvented differently ± by dividing the crack-tip
plastic zone into small elements and superposing their Weibull contributions [71].
Using the nonlocal Weibull theory, one can show that the proper statistical generalizations
of (10), with rB ˆ 0 and Eq. (12) having the correct asymptotic forms for D ! 1; D ! 0 and
m ! 1, are:
2rnd 1 D
Case …a†: rN ˆ Br0 …b m ‡ br †ÿ 2r ; bˆ ; …21†
D0
nd rnd 1 Db
Case …b†: rN ˆ r0 f m …1 ‡ rf1 ÿ †;
m r fˆ ; …22†
D
(Fig. 7) where it is assumed that rnd < m, which is normally the case.
The ®rst formula, which was obtained for r ˆ 1 in [31] and re®ned for r 6ˆ 1 by Planas, has
the property that the statistical in¯uence on the size effect disappears asymptotically for large
D. The reason is that, for long cracks or notches with stress singularity, a signi®cant contri-
bution to the Weibull probability integral comes only from the FPZ, whose size does not vary
much with D. The second formula has the property that the statistical in¯uence asymptotically
disappears for small sizes. The reason is that the FPZ occupies much of the structure volume.
Numerical analyses of test data for concrete show that the size ranges in which the statistical
in¯uence on the size effect in case (a) as well as (b) would be signi®cant do not lie within the range
of practical interest except for unreinforced beams many meters thick (arch dams, foundation
plinths). Thus, the deterministic size effect dominates and its statistical correction in (21) and

Fig. 7. Scaling laws according to nonlocal general-

ization of Weibull theory for failures after long
stable crack growth (a) or at crack initiation (b)
 (22) may be ignored for normal concrete structures, with the exception of the rare situations
where the deterministic size effect vanishes (e.g. for centric tension of an unreinforced bar).
8
Other size effects

8.1
Hypothesis of fractal origin of size effect
The partly fractal nature of crack surfaces and of the distribution of microcracks in concrete
has recently been advanced as the physical origin of the size effects observed on concrete
structures. A possible role of fractality in size effects in sea ice was discussed in [33]. In [38±40]
720 the so-called `multi-fractal scaling law' (MFSL) was proposed for failures occurring at fracture
initiation from a smooth surface; it reads
r
A2
rN ˆ A1 ‡ ; …23†
D
where A1 ; A2 ˆ const. There are, however, four objections to the fractal theory [13]:
(i) a mechanical analysis (of either invasive or lacunar fractals) predicts a different size effect
trend than (23), disagreeing with experimental observations;
(ii) the fractality of the ®nal fracture surface should not matter, because typically about 99% of
energy is dissipated by microcracks and frictional slips on the sides of this surface;
(iii) the fractal theory does not predict how A1 and A2 should depend on the geometry of the
structure, which makes the MFSL not too useful for design application;
(iv) the MFSL is a special case of the second formula in (12) for r ˆ 2, which logically follows
from fracture mechanics:
EGf EGf g 00 …0†
A1 ˆ ; A2 ˆ ÿ ; …24†
cf g 0 …0† 2cf ‰g 0 …0†Š3

[7]. Unlike fractality, the fracture explanation of (23) has the advantage that, by virtue of these
formulae, the geometry dependence of the size effect coef®cients can be determined.
8.2
Boundary layer, singularity, and diffusion
Aside from the statistical and quasibrittle size effects, there are three further types of size effect
that in¯uence nominal strength:
1. The boundary layer effect, which is due to material heterogeneity (i.e. the fact that the
surface layer of heterogeneous material such as concrete has a different composition because
the aggregates cannot protrude through the surface), and to Poisson effect (i.e. the fact that a
plane strain state on planes parallel to the surface can exist in the core of the test specimen
but not at its surface).
2. The existence of a three-dimensional stress singularity at the intersection of crack edge with
a surface, which is also caused by the Poisson effect ([28], Sec. 1.3). This causes the portion
of the FPZ near the surface to behave differently from that in the interior.
3. The time-dependent size effects caused by diffusion phenomena such as the transport of
heat or the transport of moisture and chemical agents in porous solids (this is manifested,
e.g., in the effect of size on shrinkage and drying creep, due to size dependence of the drying
half-time [22] and its effect on shrinkage cracking [90].

9
Closing remarks
Substantial though the recent progress has been, the understanding of the scaling problems of
solid mechanics is nevertheless far from complete. Mastering the size effect that bridges different
behaviors on adjacent scales in the microstructure of material will be contingent upon the de-
velopment of realistic material models that possess a material (or characteristic) length. The
theory of nonlocal continuum damage will have to move beyond the present phenomenological
approach based on isotropic spatial averaging, and take into account the directional and tensorial
interactions between the effects causing nonlocality. A statistical description of such interactions
will have to be developed. Discrete element models of the microstructure of fracturing or
damaging materials will be needed to shed more light on the mechanics of what is actually
happening inside the material and to separate the important processes from the unimportant
ones.

Appendix:
Broad-range size effect law
The simple size effect law in (10) is normally adequate for a size range up to about 1:20. For a
size range up to about 1:500, the generalized size effect law (11) (with rR = 10) has been shown
to give excellent approximation of the numerical results obtained by Hillerborg with the co-
hesive crack model for the range 1:250 (BazÆant 1985). However, only the ®rst term of the
asymptotic expansion in terms of 1/D is correct. For a modi®ed formula rN ˆ rP …1 ‡ rb†ÿ1=2r , 721
the ®rst two terms in 1/D are correct but the description of size effect in the intermediate range,
which is more important, is poor.
The following three broad-range size effect laws (proposed here), which are capable of
approximating the large-size asymptotic behavior (ensuing for example from the cohesive
crack model) up to order n ‡ 2 in 1/D, are currently investigated:

r20 XN
q q
X
N
r2N ˆ hn …1 ÿ eÿDn =D †; hn ˆ 1 …25†
1 ‡ D=D0 nˆ1 nˆ1

r20 XN
hn Dn
q
r2N ˆ …26†
1 ‡ D=D0 nˆ1 Dqn ‡ Dq
     s
2 r20 1 ‡ D=H1 1 ‡ D=H2 1 ‡ D=HN EGf
rN ˆ


; r0 ˆ 0
…27†
1 ‡ D=D0 1 ‡ D=D1 1 ‡ D=D2 1 ‡ D=DN g …a†cf

Here D1 ; . . . ; DN , H1 ; . . . ; HN , h1 ; . . . ; hN , D1 ; . . . ; DN and q are positive constants. To ensure that

each term of the chain product in (3) would not increase, it is necessary that Hi  Di
…i ˆ 1; 2; . . . ; N†. It is evident, that these formulae preserve the ®nite limit lim rN ˆ r0 for D ! 0.
Progressively increasing the number of terms in these formulae, one obtains the `cascading'
size effect plot shown in Fig. 8, in which the individual terms of (25)±(27) terminate with
progressively shifted asymptotes of slope )1/2. Each asymptote corresponds to a progressively
higher fracture energy Gf ;i . Obtaining very accurate rN values, e.g., with the cohesive crack
model, for a broad range such as 1:10000, and ®tting these results with one of these three
formulae, one could thus obtain a set of Gf ;i values, which can be regarded as a fracture energy
spectrum of the cohesive crack model de®ned by a certain softening stress-displacement curve.
The ®rst fracture energy, Gf ˆ Gf ;1 , corresponds to the area under the initial tangent of the
softening curve. This is known to suf®ce for predicting with the cohesive crack model the
maximum loads of structures for a range of about 1:20 (which covers most practical cases). If
the size range is extended much farther, say as 1:200, then much of the area under the softening
curve of the cohesive crack model, except for a remote tail of the curve, must matter for the
maximum load predictions. This area may then be associated with the fracture energy Gf ;2 ,

Fig. 8. Concept of cascading size effect for a very

broad size range
 which ought to coincide with the fracture energy GF determined by the work-of-fracture
method introduced for concrete by Hillerborg. It is well known that GF is about 2 to 2.5 times
larger than Gf . This apparent discrepancy can thus be explained in terms of the broad-range
size effect law.
In analogy to what is known from the theory of retardation spectrum in linear viscoelasticity,
one may expect the problem of identi®cation of Di from a given size effect curve to be ill-
conditioned, which means that these values would need to be chosen, within some constraints,
just as the retardation times of a viscoelastic material with a broad retardation spectrum must
be chosen. Therefore, the discrete fracture energy spectrum Gf ;1 ; Gf ;2 ; . . . ; Gf ;N will be non-
unique, depending on the choice of D0 ; D1 ; . . . ; DN in the foregoing formulae. Similar to the
continuous retardation spectrum in viscoelasticity, one could consider in®nitely closely spaced
722
Di values and generalize (25) to a continuous spectrum of fracture energies, which would then
be unique.

References
1. Argon, A. S.: Fracture of composites. Treatise of Materials Science and Technology, vol. 1. p. 79. New
York: Academic Press 1972
2. Barenblatt, G. I.: The formation of equilibrium cracks during brittle fracture. General ideas and
hypothesis, axially symmetric cracks. Prikl. Mat. Mekh. 23(3) (1959) 434±444
3. Barenblatt, G. I.: The mathematical theory of equilibrium cracks in brittle fracture. Advanced Appl.
Mech. 7 (1962) 55±129
4. Barenblatt, G. I.: Similarity, self-similarity and intermediate asymptotics. Consultants Bureau, New
York 1979
5. BazÏant, Z. P.: (1976) Instability, ductility, and size effect in strain-softening concrete. J. Eng. Mech.
Div., Am. Soc. Civil Engrs., 102, EM2, 331±344; disc. 103, 357±358, 775±777, 104, 501±502
6. BazÏant, Z. P.: Size effect in blunt fracture: Concrete, rock, metal. J. Eng. Mech. ASCE 110 (1984) 518±
535
7. BazÏant, Z. P.: Size effect in tensile and compression fracture of concrete structures: computational
modeling and design. In: Mihashi, H.; Rokugo, K.: (eds.) Fracture Mechanics of Concrete Structures.
3rd Int. Conf., FraMCoS-3(Gifu, Japan) pp. 1905±1922, Freiburg, Aedi®catio Publishers 1998
8. BazÏant, Z. P.: Fracture Mechanics of Concrete Structures. Proc. First Intern. Conf. (FraMCoS-1,
Breckenridge). p. 1040, June 1±5, London: Elsevier 1992
9. BazÏant, Z. P.: Large-scale thermal bending fracture of sea ice plates. J. Geophys. Res. 97 (C11) (1992)
17,739±17,751
10. BazÏant, Z. P.: Scaling laws in mechanics of failure. J. Eng. Mech. ASCE 119(9) (1993) 1828±1844
11. BazÏant, Z. P.: Fracturing truss model: Size effect in shear failure of reinforced concrete. J. Eng. Mech.
ASCE 123(12) (1997) 1276±1288
12. BazÏant, Z. P.: Scaling of quasibrittle fracture: asymptotic analysis. Int. J. Fracture 83(1) (1997) 19±40
13. BazÏant, Z. P.: Scaling of quasibrittle fracture: hypotheses of invasive and lacunar fractality, their
critique and Weibull connection. Int. J. Fracture 83(1) (1997) 41±65
14. BazÏant, Z. P.: Stability of structures. ASME Appl. Mech. Rev. (submitted)
15. BazÏant, Z. P.; Cedolin, L.: Stability of Structures: Elastic, Inelastic, Fracture and Damage Theories
(textbook and reference volume). New York: Oxford University Press 1991
16. BazÏant, Z. P.; Chen, E.-P.: Scaling of structural failure. Appl. Mech. Rev. ASME 50(10) (1997) 593±627
17. BazÏant, Z. P.; Daniel, I. M.; Li, Zhengzhi: Size effect and fracture characteristics of composite
laminates. ASME J. Eng. Mat. Tech. 118(3) (1996) 317±324
18. BazÏant, Z. P.; Kazemi, M. T.: Size effect in fracture of ceramics and its use to determine fracture
energy and effective process zone length. J. American Ceramic Society 73(7) (1990) 1841±1853
19. BazÏant, Z. P.; Kazemi, M. T.: Size effect on diagonal shear failure of beams without stirrups. ACI
Structural Journal 88(3) (1991) 268±276
20. BazÏant, Z. P.; Kim, J.-J. H.; Daniel, I. M.; Becq-Giraudon, E.; Zi, G.: Size effect on compression
strength of ®ber composites failing by kink band propagation. Int. J. Fracture 95 (1999), 103±141.
Special Issue on Fracture Scaling, BazÏant, Z. P.; Rajapakse, Y. D. S.: (eds.). (also in Ref. 30)
21. BazÏant, Z. P.; Kim, J.-J. H.: Size effect in penetration of sea ice plate with partthrough cracks. I.
Theory. J. Eng. Mech. ASCE 124(12) (1998) 1310±1315; ``II. Results'', ibid., 1316±1324
22. BazÏant, Z. P.; Kim, J.-J. H.: Daniel, I. M.; Becq-Giraudon, E.; Zi, G.: Size effect on compression
strength of ®ber composites failing by kink band propagation. Int. J. Fracture (in press)
23. BazÏant, Z. P.; Li, Yuan-Neng: Stability of cohesive crack model: Part I ± Energy principles. Trans.
ASME J. Appl. Mech. 62 (1995) 959±964; Part II ± Eigenvalue analysis of size effect on strength and
ductility of structures. ibid. 62 (1995) 965±969
24. BazÏant, Z. P.; Li, Yuan-Neng.: Cohesive crack with rate-dependent opening and viscoelasticity: I.
Mathematical model and scaling. Int. J. Fracture 86(3) (1997) 247±265
25. BazÏant, Z. P.; Lin, F.-B.; Lippmann, H.: Fracture energy release and size effect in borehole breakout.
Int. J. Numer. Anal. Methods Geomech. 17 (1993) 1±14
26. BazÏant, Z. P.; Oh B.-H.: Crack band theory for fracture of concrete. Mat. Struct. (RILEM, Paris) 16
(1983) 155±177
27. BazÏant, Z. P.; Pfeiffer, P. A.: Determination of fracture energy from size effect and brittleness
number. ACI Materials J. 84 (1987) 463±480
28. BazÏant, Z. P.; Planas, J.: Fracture and Size Effect in Concrete and Other Quasibrittle Materials. Boca
Raton: CRC Press 1998
29. BazÏant, Z. P.; Pfeiffer, P. A. Determination of fracture energy from size effect and brittleness number.
ACI Materials Jour. 84 (1987) 463±480
30. BazÏant, Z. P.; Rajapakse, Y. D. S. (eds.): Fracture Scaling. Dordrecht: Kluwer Academic Publishers;
also Special Issue of Int. J. Fracture 95 (1999), 1±435
31. BazÏant, Z. P.; Xi, Y.: Statistical size effect in quasi-brittle structures: II. Nonlocal theory. ASCE J. Eng.
Mech. 117(11) (1991) 2623±2640
32. Beremin, F. M.: A local criterion for cleavage fracture of a nuclear pressure vessel steel. Metallurgy
Transactions A 14 (1983) 2277±2287
33. Bhat, S. U.: Modeling of size effect in ice mechanics using fractal concepts. J. Offshore Mech. Arctic 723
Eng. 112 (1990) 370±376
34. Budiansky, B.: Micromechanics. Comput. Struct. 16(1±4) (1983) 3±12
35. Budiansky, B.; Fleck, N. A.; Amazigo, J. C.: On kink-band propagation in ®ber composites. J. Mech.
Phys. Solids 46(9) (1997) 1637±1635
36. Carpinteri, A.: Mechanical damage and crack growth in concrete. Dordrecht±Boston: M. Nijhoff Publ. ±
Kluwer (1986)
37. Carpinteri, A.: Decrease of apparent tensile and bending strength with specimen size: Two different
explanations based on fracture mechanics. Int. J. Solids Struct. 25(4) (1989) 407±429
38. Carpinteri, A.: Fractal nature of material microstructure and size effects on apparent mechanical
properties. Mech. Mat. 18 (1994) 89±101
39. Carpinteri, A.: Scaling laws and renormalization groups for strength and toughness of disordered
materials. Int. J. Solids Struct. 31 (1994) 291±302
40. Carpinteri, A.; Chiaia, B.; Ferro, G.: Multifractal scaling law for the nominal strength variation of
concrete structures. In: Mihashi, M.; Okamura, H.; BazÏant, Z. P.: (eds.) Size effect in concrete
structures (Proc., Jap. Concrete Inst. Int. Workshop, Sendai, Japan, 1993). pp. 193±206, London±New
York: E & FN Spon 1994
41. Carpinteri, A.; Chiaia, B.: Fracture Mechanics of Concrete Structures. In: Wittmann, F. H.: (ed.)
Proceedings of FraMCoS-2, E.T.H., ZuÈrich. pp. 581±596, Aedi®catio Publishers, Freiburg 1995
42. Carter, B. C.: Size and stress gradient effects on fracture around cavities. Rock Mech. Rock Eng. 25(3)
(1992) 167±186
43. Carter, B. C.; Lajtai, E. Z.; Yuan, Y.: Tensile fracture from circular cavities loaded in compression. Int.
J. Fracture 57 (1992) 221±236
Ï ervenka, V.; Pukl, R.: SBETA analysis of size effect in concrete structures. In: Mihashi, H.; Okamura,
44. C
H.; BazÏant, Z. P.: (eds.) Size Effect in Concrete Structure. pp. 323±333 London: E & FN Spon 1994
45. Cottrell, A. H.: Iron and Steel Institute Special Report 69 (1963) 281
46. Dempsey, J. P.; Adamson, R. M.; Mulmule, S. V.: Large-scale in-situ fracture of ice. In: Wittmann, F.
H. (ed.) Vol. I (Proc., 2nd Int. Conf. on Fracture Mech. of Concrete Structures (FraMCoS-2), ETH,
ZuÈrich). pp. 575±684 Freiburg: Aedi®catio Publishers 1995
47. Dempsey, J. P.; Adamson, R. M.; Mulmule, S. V.: Scale effect on the in-situ tensile strength and failure
of ®rst-year sea ice at Resolute, NWR. Int. J. Fracture (Special Issue on Fracture Scaling, BazÏant, Z. P.;
Rajapakse, Y. D. S.: (eds.)) 95 (1999) 325±345
48. Dempsey, J. P.; Slepyan, L. I.; Shekhtman, I. I.: Radial cracking with closure. Int. J. Fracture 73(3)
(1995) 233±261
49. Dugdale, D. S.: Yielding of steel sheets containing slits. J. Mech. Phys. Solids 8 (1960) 100±108
50. Evans, A. G.: A general approach for the statistical analysis of multiaxial fracture. J. of the American
Ceramic Soc. 61 (1978) 302±308
51. FreÂchet, M.: Sur la loi de probabilite de l' eÂcart maximum. Ann. Soc. Polon. Math. 6 (1927) 93
52. Fischer, R. A.; Tippett, L. H. C.: Limiting forms of the frequency distribution of the largest and
smallest member of a sample. Proc., Cambridge Philosophical Society 24 (1928) 180±190
53. Frankenstein, E. G.: Load test data for lake ice sheet. Technical Report 89, U.S. Army Cold Regions
Research and Engineering Laboratory. Hannover-New Hampshire 1963
54. Frankenstein, E. G.: Stength of ice sheets. Proc., Conf. on Ice Pressures against Struct.; Tech. Memor. No.
92, NRCC No. 9851. pp. 79±87. Laval University Quebec, National Research Council of Canada 1966
55. Freudenthal, A. M.: Physical and statistical aspects of fatigue. Advance in Appl. Mech. 4 (1956) 117±157
56. Freudenthal, A. M.: Statistical approach to brittle fracture. In: Liebowitz, H.: (ed.) Fracture, vol. 2
Chapter 6, pp. 591±619. Academic Press 1968
57. Freudenthal, A. M.; Gumbell, E. J.: Physical and statistical aspects of fatigue. Advances in Appl. Mech.
4 (1956) 117±157
58. Galileo, Galilei Linceo: Discorsi i Demostrazioni Matematiche intorno aÁ due Nuove Scienze, Elsevirii,
Leiden 1638. pp. 178±181 (English transl. by T. Weston, London 1730)
59. Gettu, R.; BazÏant, Z. P.; Karr, M. E.: Fracture properties and brittleness of high-strength concrete.
ACI Mat. J. 87 (1990) 608±618
60. Grif®th, A. A.: The phenomena of rupture and ¯ow in solids. Phil. Trans. 221A (1921) 179±180
61. Haimson, B. C.; Herrick, C. G.: In-situ stress calculation from borehole breakout experimental
studies. Proc., 26th U.S. Symp. Rock Mech. (1989) 1207±1218
 62. Hillerborg, A.; ModeÂer, M.; Petersson, P. E.: Analysis of crack formation and crack growth in concrete
by means of fracture mechanics and ®nite elements. Cement and Concrete Research 6 (1976) 773±782
63. Iguro, M.; Shiyoa, T.; Nojiri, Y.; Akiyama, H.: Experimental studies on shear strength of large
reinforced concrete beams under uniformly distributed load. Concrete Library International, Japan
Soc. Civil Engrs. 5 (1985) 137±154 (translation of 1984 article in Proc. JSCE)
64. Jenq, Y. S.; Shah, S. P.: A two parameter fracture model for concrete. J. Eng. Mech. ASCE 111(4)
(1985) 1227±1241
65. Kani, G. N. J.: Basic facts concerning shear failure. ACI J. Proc. 64 (1967) 128±141
66. Kaplan, M. F.: Crack propagation and the fracture concrete. ACI J. V. 58(11) (1961)
67. Kerr, A. D.: Bearing capacity of ¯oating ice covers subjected to static, moving, and oscillatory loads.
Appl. Mech. Rev. ASME 49(11) (1996) 463±476
68. Kesler, C. E.; Naus, D. J.; Lott, J. L.: Fracture Mechanics ± Its applicability to concrete. In: Proc. Int.
724 Conf. Mechanical Behavior of Materials. Vol. IV, 1972, pp. 113±124. The Soc. Mater. Sci. Kyoto:
1971
69. Kittl, P.; Diaz, G.: Weibull's fracture statistics, or probabilistic strength of materials: state of the art.
Res Mechanica 24 (1988) 99±207
70. Kittl, P.; Diaz, G.: Size effect on fracture strength in the probabilistic strength of materials. Reliability
Eng. Sys. Saf. 28 (1990) 9±21
71. Lei, Y.; O'Dowd, N. P.; Busso, E. P.; Webster, G. A.: Weibull stress solutions for 2-D cracks in elastic
and elastic-plastic materials. Int. J. Fracture 89 (1998) 245±268
72. Leicester, R. H.: The size effect of notches. Proc., 2nd Australasian Conf. Mech. Struct. Mater., pp.
4.1±4.20 1969 Melbourne
73. Li, Yuan-Neng; BazÏant, Z. P.: Cohesive crack with rate-dependent opening and viscoelasticity: II.
numerical algorithm, behavior and size effect. Int. J. Fracture 86(3) (1997) 267±288
74. Li, Zhengzhi; BazÏant, Z. P.: Acoustic emissions in fracturing sea ice plate simulated by particle
system. J. Eng. Mechanics ASCE 124(1) (1998) 69±79
75. Lichtenberger, G. J.; Jones, J. W.; Stegall, R. D.; Zadow, D. W.: Static ice loading tests: Resolute Bay ±
Winter 1973/74. APOA Project No. 64, Rep. No. 745B-74-14 (CREEL Bib. No. 34-3095). Sunoco Sci.
Technol., Rechardson, Texas (1974)
76. Mariotte, E.: Traite du movement des eaux. Posthumously edited by M. de la Hire. Engl. transl. by
J. T. Desvaguliers, London 1718. p. 249, also Mariotte's collected works, 2nd edn. The Hague 1740
77. Marti, P.: Size effect in double-punch tests on concrete cylinders. ACI Materials J. 86 No. 6 (1989)
597±601
78. Mihashi, H.: Stochastic theory for fracture of concrete. In: Wittmann, F. H.: (ed.) Fracture mechanics
of concrete. (1983) pp. 301±339. Amsterdam: Elsevier Science Publishers, B.V.
79. Mihashi, H.; Izumi, M.: Stochastic theory for concrete fracture. Cem. Concr. Res. 7 (1977) 411±422
80. Mihashi, H.; Okamura, H.; BazÆant, Z. P. (eds.): Size effect in concrete structures (Proc., Japan
Concrete Institute Intern. Workshop in Sendai, Japan Oct. 31±Nov.2 1995). London-New York: E &
FN Spon 1994
81. Mihashi, H.; Rokugo, K. (eds.): Fracture Mechanics of Concrete Structures (Proc., 3rd Int. Conf.,
FraMCoS-3 held in Gifu, Japan) Freiburg: Aedi®catio Publishers 1998
82. Mihashi, H.; Zaitsev, J. W.: Statistical nature of crack propagation. In: Wittmann, F. H.: (ed.) Section
4-2 in Report to RILEM TC 50-FMC. 1981
83. Mulmule, S. V.; Dempsey, J. P.; Adamson, R. M.: Large-scale in-situ ice fracture experiments±Part II:
Modeling efforts. Ice Mechanics±1995 (ASME Joint Appl. Mechanics and Materials Summer Conf.,
held at University of California AMD-MD '95, New York: Am. Soc. of Mech. Engrs. 1995
84. Nesetova, V.; Lajtai, E. Z.: Fracture from compressive stress concentration around elastic ¯aws. Int. J.
Rock Mech. Mining Sci. 10 (1973) 265±284
85. Okamura, H.; Maekawa, K.: Experimental study of Size effect in concrete structures, in size effect in
concrete structures. pp. 3±24 In: Mihashi, H.; Okamura, H.; BazÆant, Z. P.: (eds.) Proc. of JCI Intern.
Workshop held in Sendai, Japan, 1993) London E & FN Spon 1994
86. Peirce, F. T.: J. Textile Inst. 17 (1926) 355
87. Petersson, P. E.: Crack growth and development of fracture zones in plain concrete and similar
materials. Report TVBM-1006, Div. of Building Materials, Lund Inst. of Tech., Sweden (1981)
88. Planas, J.; Elices, M.: Conceptual and experimental problems in the determination of the fracture
energy of concrete. Proc. Int. Workshop on Fracture Toughness and Fracture Energy, Test Methods
for Concrete and Rock. pp. 203±212. Tohoku Univ. Sendai, Japan 1988
89. Planas, J.; Elices, M.: Cracking and Damage. In: Mazars, J.; BazÏant, Z. P.: (eds.) pp. 62±476 London:
Elsevier 1989
90. Planas, J.; Elices, M.: Drying shrinkage effects on the modulus of rupture. Creep and Shrinkage of
Concrete (Proc., 5th Int. RILEM Symp., Barcelona) BazÏant, Z. P.; Carol, I.: (eds.) pp. 357±368.
London: E & FN Spon 1993
91. Planas, J.; Elices, M.; Guinea, G. V.: Cohesive cracks vs. nonlocal models: Closing the gap. Int. J.
Fracture 63(2) (1993) 173±187
92. Reinhardt, H. W.: Masstabsein¯uss bei Schubversuchen im Licht der Bruchmechanik. Beton und
Stahlbetonbau 1 (1981) 19±21
93. RILEM Recommendation Size effect method for determining fracture energy and process zone of
concrete. Mat. Struct. 23 (1990) 461±465
94. Rosen, B. W.: Mechanics of composite strengthening. Fiber Composite Materials. Chapter 3, Metals
Park, Ohio: pp. 37±75 American Soc. for Metals Seminar, Am Soc. for Metals 1965
95. Ruggieri, C.; Dodds, R. H.: Transferability model for brittle fracture including constraint and ductile
tearing effects ± a probabilistic approach. Int. J. Fracture 79 (1996) 309±340
96. Sedov, L. I.: Similarity and dimensional methods in mechanics. New York: Academic Press 1959
97. Selected Papers by Alfred M. Freudenthal New York: Am. Soc. Civil Engrs. 1981
98. Shioya, Y.; Akiyama, H.: (1994) Application to design of size effect in reinforced concrete structures.
In: Mihashi, H.; Okamura, H.; BazÏant, Z. P.: (eds.) Size Effect in Concrete Structures (Proc. Intern.
Workshop of Sendai, 1993) London: E & FN Spon 409±416
99. Slepyan, L. I.: Modeling of fracture of sheet ice. Izvestia AN SSSR, Mekh, Tverd. Tela 25(2) (1990)
151±157
100. Sodhi, D. S.: Breakthrough loads of ¯oating ice sheets. J. Cold Regions Eng. ASCE 9(1) (1995) 4±20
101. Tippett, L. H. C.: On the extreme individuals and the range of samples. Biometrika 17 (1925) 364 725
102. von Mises, R.: La distribution de la plus grande de n valeurs. Rev. Math. Union Interbalcanique 1 (1936) 1
103. Walraven, J.; Lehwalter: Size effects in short beams loaded in shear. ACI Struct. J. 91(5) (1994) 585±593
104. Walraven, J.: Size effects: their nature and their recognition in building codes. Studie Ricerche
(Politecnico di Milano) 16 (1995) 113±134
105. Walsh, P. F.: Fracture of plain concrete. Indian Concrete J. 46(11) (1972)
106. Walsh, P. F.: Crack initiation in plain concrete. Magazine of Concrete Research 28 (1976) 37±41
107. Weibull, W.: The phenomenon of rupture in solids. Proc. Royal Swedish Inst. Eng. Res. (Ing-
enioersvetenskaps Akad. Handl.) 153 (1939) 1±55
108. Weibull, W.: A statistical representation of fatigue failures in solids. Proc. Roy. Inst. Techn. No. 27 1949
109. Weibull, W.: A statistical distribution function of wide applicability. J. Appl. Mech. ASME 18 (1951)
110. Weibull, W.: Basic aspects of fatigue. Proc. Colloquium on Fatigue. Springer-Verlag: Stockholm 1956
111. Wells, A. A.: Unstable crack propagation in metals-cleavage and fast fracture. Symp. on Crack
Propagation, Cran®eld 1 (1961) 210±230
112. Williams, E.: Some observations of Leonardo, Galileo, Mariotte and others relative to size effect.
Annals of Science 13 (1957) 23±29; da Vinci, L. (1500's) ± see The Notebooks of Leonardo da Vinci
(1945) Edward McCurdy, London (p. 546); and Les Manuscrits de LeÂonard de Vinci, transl. in French
by C. Ravaisson-Mollien, Institut de France (1881±91), Vol. 3
113. Wisnom, M. R.: The relationship between tensile and ¯exural strength of unidirectional composite.
J Composite Mat. 26(8) (1992) 1173±1180
114. Wittmann, F. H.: (ed.) Fracture Mechanics of Concrete Structures (Proc., 2nd Int. Conf. on Fracture
Mech. of Concrete and Concrete Structures (FraMCoS-2) ETH, ZuÈrich) pp. 515±534. Freiburg: Ae-
di®catio Publishers 1995
115. Wittmann, F. H.; Zaitsev, Yu., V.: Crack propagation and fracture of composite materials such as
concrete. Proc. 5th Int. Conf. on Fracture (ICF5), Cannes 1981
116. Zaitsev, J. W.; Wittmann, F. H.: A statistical approach to the study of the mechanical behavior of
porous materials under multiaxial state of stress. Proc. of the 1973 Symp. on Mechanical Behavior of
Materials (Kyoto, Japan) (1974) 705
117. Zech, B.; Wittmann, F. H.: A complex study on the reliability assessment of the containment of a
PWR, Part II. Probabilistic approach to describe the behavior of materials. In: Jaeger, T. A.; Boley, B.
A.: (eds.) Trans. 4th Int. Conf. on Structural Mechanics in Reactor Technology, Vol. H, J1/11, 1±14
1977 Brussels: European Communities